why not average results (for stock) geekbench 4 under the same OS?
[....]
I'm confident in saying that for normal use the i3 2130 would be mostly superior.
The OS should not matter much (unless running 32bit verus 64bit which). I mainly wanted to avoid all the overclocked results and outliers (of which in GB often there can be many). This i3 is luckily not unlocked, so the OC potential is less; but people can still do base clock overclock, or put in a monster cooler (changing base freq to all core boost freq). I know Macs are impossible to hard to overclock.
The Athlon 845 lacks an L3 cache (and the L2 shrunk in size by half from the previous generation--although with a better faster low latency L2 cache). So a tip for Athlon 845 builders is match it with low latency memory (low cas/freq ) if performance is a bigger priority than budget. Also it seems to be true that A88 mobos seem to do better; all the top results are A88 vs A68. But since perf is secondary to budget for this category, the difference likely isn't worth the extra cost for the typical person.
--
Have you ever tried a desktop excavator quadcore (or even any well binned >= 15w excavator quad)? For those who haven't, they're worlds different from early gen dozers.
--
The reason for GB3 over GB4 is that they break down performance by FP and INT tasks. It gives the layman (who is not very familiar with all the particulars of a large slew of benchmarks) a better idea of what the benchmarks are, and also gives you not only a single total composite, but a composite for all four categories I listed above (int, int-mul, fp, fp-mul).
There's only one thing I wish GB3 did, which would be mixed load (fp+int) for various numbers of threads: 2, 4, 6, 8. That would give people a realistic idea of typical real world performance.
Cinebench is not a bad benchmark, but it seems like mostly pure floats loads to me (maybe I'm wrong). Unless you are some niche user, it doesn't seem like a typical real life situation.